Machine-Learning & Data-driven identification of variants to improve neonatal care and patient outcomes

Summary

This project exploits the population data held in the National Neonatal Research Database (NNRD), to develop and apply machine-learning (ML) techniques for the systematic identification of unwarranted variation in a set of clinical outcomes, and their principal care-related determinants.

Collaborators: The Neonatal Data Analysis Unit at the Chelsea and Westiminster Hospital, with Prof Neena Modi, Dr. Chris Gale, Dr. Cheryl Battersby and Kayleigh Ougham.

This collaboration has been funded by the Imperial BRC, the  Chelsea and Westiminster Foundation and a MRC Partnership grant.

This is the first time ML tools are applied to the NNRD cohort. Our initial application has focused on discovery of feeding patterns for very pre-term babies and their association with outcomes such as mortality, length of stay or breast-milk feeding at discharge. We are now looking at growth patterns over 12 years, use of deep-learning to encode treatment patterns and "health status", modeling of care pathways, and application of FAIR AI principles on sensitive variables. 

 

Contact Group member(s): Sam Greenbury, Elsa Angelini (previous members: Jinyi Wu)

 


 

Contact Group member(s):

Paul Blakeley